Dreaming Species

project description:Dreaming Species is an online listening experience designed for headphones, conceived and created by artist composer Melanie Wilson (U.K.). Across three episodes, this dream-like piece embarks on a journey into a world in which human connection to animals is re-imagined using the processes of machine learning. To engage with the crisis of biodiversity loss in the UK, Wilson collaborated with an AI neural net, using the voices of animals at risk of extinction and human voices, creating a new sonic language. Combining multi-part vocal music, animal and machine voice, text and binaural recording techniques, Dreaming Species is the opening creative chapter of a multi-year research process, in which Wilson has explored the possibilities of machine learning to imagine new sonic meeting grounds for human and animal narratives. Dreaming Species can be experienced online at: www.dreamingspecies.com background:Dreaming Species is part of Wilson’s ongoing work in exploring the shared processes of human memory and machine learning to deeply listen to our more than human companions.Wilson has collaborated with PRiSM (Practice Research in Science and Music) at Royal Northern College of Music, Manchester, U.K. to create the machine generated voice in this piece. This unique hybrid ‘voice’ has been created through iterative training of the SampleRNN neural net, using a corpus of human and animal vocality. PRiSM developed an open-source software tool, through modernising and reimplementing the pre-existing SampleRNN architecture - a specific type of Recurrent Neural Network (RNN) for generating raw audio samples based upon learning any custom corpus of sound or music. Released in early 2020, PRiSM SampleRNN has opened many new avenues for composing and producing music - perfect for working with actual recordings of animal and human voice.The project is funded by Sound and Music and Arts Council England. www.melaniewilson.org.uk